Abarbanell and Lehavy (2003a) empirically link abnormal accruals to two asymmetries in the forecast error distribution (i.e., a high incidence of extreme negative forecast errors and a high incidence of small positive errors) and identify two alternative explanations: (1) that analyst bias is differentially exacerbated by abnormal accruals, or (2) managerial discretion introduces a bias in reported earnings that analysts are unable or unwilling to incorporate in their forecasts. For the lower tail asymmetry, we test these competing hypotheses using forward and reverse regressions that relate contemporaneous forecasts and reported earnings conditional on various earnings components (i.e., cash flow, normal and abnormal accruals). We fail to find that analyst forecast bias is exacerbated by large abnormal accruals. However, our evidence is consistent with analysts being unable or unwilling to forecast managerial discretion in reported earnings, particularly non-current abnormal accruals, which contributes to the frequency of large negative forecast errors. Alternatively, applying the methodology in Abarbanell and Lehavy (2003a) to the middle asymmetry issue, we show that abnormal accruals do not uniquely contribute to firms just beating analysts' forecasts. Further, using an alternative methodology, we fail to find that abnormal accruals contribute to the middle asymmetry. Rather, we find evidence that both income-increasing real earnings management activities and analyst inefficiencies are likely contributors to the middle asymmetry phenomenon.